ZFS cannot promote: not a cloned filesystem

patricoferris commented 3 years ago

An issue to track the current bug we are seeing on the Github runner with ZFS where during the stress test we get the following error:

+ opam exec -- dune exec -- ./stress/stress.exe zfs:zfs
cannot promote 'zfs/cache-tmp/0-c-c1': not a cloned filesystem
stress: internal error, uncaught exception:
        (Failure
          "\"sudo\" \"zfs\" \"promote\" \"zfs/cache-tmp/0-c-c1\" failed with exit status 1")

Note, previously on Travis builds this was working (with the same version of ZFS): https://travis-ci.org/github/ocurrent/obuilder/builds/768095520 and the stress test works with an older version of ZFS I've tried locally. Setting the stress test to output debug logs also suggests all of the zfs commands leading up to this call to promote are identical so I'm wondering if there is some weird github runner issue happening here. I'll see if I can update to the right version locally to see if that shows the bug.

patricoferris commented 3 years ago

Looks like that is the issue and there doesn't seem to be any answer as to how to fix it... Using a third-party ppa is not an ideal solution but it shows (I think) that this is the problem https://github.com/patricoferris/obuilder/runs/4324615404?check_suite_focus=true

Edit: which is essentially https://github.com/openzfs/zfs/issues/9894

talex5 commented 3 years ago

I think using a PPA is fine to get the CI passing. And maybe link to the issue from the error message?

ocurrent / obuilder

ZFS cannot promote: not a cloned filesystem #89